WWW Mechanize: Handy Web Browsing in A Perl Object

NAME

WWW::Mechanize - Handy web browsing in a Perl object

VERSION

version 2.15

SYNOPSIS

WWW::Mechanize supports performing a sequence of page fetches including following links and submitting forms. Each fetched page is parsed and its links and forms are extracted. A link or a form can be selected, form fields can be filled and the next page can be fetched. Mech also stores a history of the URLs you've visited, which can be queried and revisited.

use WWW::Mechanize ();
my $mech = WWW::Mechanize->new();

$mech->get( $url );

$mech->follow_link( n => 3 );
$mech->follow_link( text_regex => qr/download this/i );
$mech->follow_link( url => 'http://host.com/index.html' );

$mech->submit_form(
    form_number => 3,
    fields      => {
        username    => 'mungo',
        password    => 'lost-and-alone',
    }
);

$mech->submit_form(
    form_name => 'search',
    fields    => { query  => 'pot of gold', },
    button    => 'Search Now'
);

# Enable strict form processing to catch typos and non-existant form fields.
my $strict_mech = WWW::Mechanize->new( strict_forms => 1);

$strict_mech->get( $url );

# This method call will die, saving you lots of time looking for the bug.
$strict_mech->submit_form(
    form_number => 3,
    fields      => {
        usernaem     => 'mungo',           # typo in field name
        password     => 'lost-and-alone',
        extra_field  => 123,               # field does not exist
    }
);

DESCRIPTION

WWW::Mechanize, or Mech for short, is a Perl module for stateful programmatic web browsing, used for automating interaction with websites.

Features include:

  • All HTTP methods
  • High-level hyperlink and HTML form support, without having to parse HTML yourself
  • SSL support
  • Automatic cookies
  • Custom HTTP headers
  • Automatic handling of redirections
  • Proxies
  • HTTP authentication

Mech is well suited for use in testing web applications. If you use one of the Test::*, like Test::HTML::Lint modules, you can check the fetched content and use that as input to a test call.

use Test::More;
like( $mech->content(), qr/$expected/, "Got expected content" );

Each page fetch stores its URL in a history stack which you can traverse.

$mech->back();

If you want finer control over your page fetching, you can use these methods. [follow_link()](#mech-follow_link) and [submit_form()](#mech-submit_form) are just high level wrappers around them.

$mech->find_link( n => $number );
$mech->form_number( $number );
$mech->form_name( $name );
$mech->field( $name, $value );
$mech->set_fields( %field_values );
$mech->set_visible( @criteria );
$mech->click( $button );

WWW::Mechanize is a proper subclass of LWP::UserAgent and you can also use any of LWP::UserAgent's methods.

$mech->add_header($name => $value);

Please note that Mech does NOT support JavaScript, you need additional software for that. Please check "JavaScript" in WWW::Mechanize::FAQ for more.

IMPORTANT LINKS

https://github.com/libwww-perl/WWW-Mechanize/issues

The queue for bugs & enhancements in WWW::Mechanize. Please note that the queue at http://rt.cpan.org is no longer maintained.

https://metacpan.org/pod/WWW::Mechanize

The CPAN documentation page for Mechanize.

https://metacpan.org/pod/distribution/WWW-Mechanize/lib/WWW/Mechanize/FAQ.pod

Frequently asked questions. Make sure you read here FIRST.

CONSTRUCTOR AND STARTUP

new()

Creates and returns a new WWW::Mechanize object, hereafter referred to as the "agent".

my $mech = WWW::Mechanize->new()

The constructor for WWW::Mechanize overrides two of the params to the LWP::UserAgent constructor:

agent => 'WWW-Mechanize/#.##'
cookie_jar => {}    # an empty, memory-only HTTP::Cookies object

You can override these overrides by passing params to the constructor, as in:

my $mech = WWW::Mechanize->new( agent => 'wonderbot 1.01' );

If you want none of the overhead of a cookie jar, or don't want your bot accepting cookies, you have to explicitly disallow it, like so:

my $mech = WWW::Mechanize->new( cookie_jar => undef );

Here are the params that WWW::Mechanize recognizes. These do not include params that LWP::UserAgent recognizes.

autocheck => [0|1]

Checks each request made to see if it was successful. This saves you the trouble of manually checking yourself. Any errors found are errors, not warnings.

The default value is ON, unless it's being subclassed, in which case it is OFF. This means that standalone WWW::Mechanize instances have autocheck turned on, which is protective for the vast majority of Mech users who don't bother checking the return value of [get()](#mech-get-uri) and [post()](#mech-post-uri-content-content) and can't figure why their code fails. However, if WWW::Mechanize is subclassed, such as for Test::WWW::Mechanize or Test::WWW::Mechanize::Catalyst, this may not be an appropriate default, so it's off.

noproxy => [0|1]

Turn off the automatic call to the LWP::UserAgent env_proxy function.

This needs to be explicitly turned off if you're using Crypt::SSLeay to access a https site via a proxy server. Note: you still need to set your HTTPS_PROXY environment variable as appropriate.

onwarn => \&func

Reference to a warn-compatible function, such as [Carp](https://metacpan.org/pod/Carp)::carp, that is called when a warning needs to be shown.

If this is set to undef, no warnings will ever be shown. However, it's probably better to use the quiet method to control that behavior.

If this value is not passed, Mech uses Carp::carp if Carp is installed, or CORE::warn if not.

onerror => \&func

Reference to a die-compatible function, such as [Carp](https://metacpan.org/pod/Carp)::croak, that is called when there's a fatal error.

If this is set to undef, no errors will ever be shown.

If this value is not passed, Mech uses Carp::croak if Carp is installed, or CORE::die if not.

quiet => [0|1]

Don't complain on warnings. Setting quiet => 1 is the same as calling $mech->quiet(1). Default is off.

stack_depth => $value

Sets the depth of the page stack that keeps track of all the downloaded pages. Default is effectively infinite stack size. If the stack is eating up your memory, then set this to a smaller number, say 5 or 10. Setting this to zero means Mech will keep no history.

In addition, WWW::Mechanize also allows you to globally enable strict and verbose mode for form handling, which is done with HTML::Form.

strict_forms => [0|1]

Globally sets the HTML::Form strict flag which causes form submission to croak if any of the passed fields don't exist in the form, and/or a value doesn't exist in a select element. This can still be disabled in individual calls to [submit_form()](#mech-submit_form).

Default is off.

verbose_forms => [0|1]

Globally sets the HTML::Form verbose flag which causes form submission to warn about any bad HTML form constructs found. This cannot be disabled later.

Default is off.

marked_sections => [0|1]

Globally sets the HTML::Parser marked sections flag which causes HTML CDATA[[ sections to be honoured. This cannot be disabled later.

Default is on.

To support forms, WWW::Mechanize's constructor pushes POST on to the agent's requests_redirectable list (see also LWP::UserAgent.)

$mech->agent_alias( $alias )

Sets the user agent string to the expanded version from a table of actual user strings. $alias can be one of the following:

  • Windows IE 6
  • Windows Mozilla
  • Mac Safari
  • Mac Mozilla
  • Linux Mozilla
  • Linux Konqueror

then it will be replaced with a more interesting one. For instance,

$mech->agent_alias( 'Windows IE 6' );

sets your User-Agent to

Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1)

The list of valid aliases can be returned from [known_agent_aliases()](#mech-known_agent_aliases). The current list is:

  • Windows IE 6
  • Windows Mozilla
  • Mac Safari
  • Mac Mozilla
  • Linux Mozilla
  • Linux Konqueror

$mech->known_agent_aliases()

Returns a list of all the agent aliases that Mech knows about. This can also be called as a package or class method.

@aliases = WWW::Mechanize::known_agent_aliases();
@aliases = WWW::Mechanize->known_agent_aliases();
@aliases = $mech->known_agent_aliases();

PAGE-FETCHING METHODS

$mech->get( $uri )

Given a URL/URI, fetches it. Returns an HTTP::Response object. $uri can be a well-formed URL string, a URI object, or a WWW::Mechanize::Link object.

The results are stored internally in the agent object, but you don't know that. Just use the accessors listed below. Poking at the internals is deprecated and subject to change in the future.

get() is a well-behaved overloaded version of the method in LWP::UserAgent. This lets you do things like

$mech->get( $uri, ':content_file' => $filename );

and you can rest assured that the params will get filtered down appropriately. See "get" in LWP::UserAgent for more details.

NOTE: Because :content_file causes the page contents to be stored in a file instead of the response object, some Mech functions that expect it to be there won't work as expected. Use with caution.

Here is a non-complete list of methods that do not work as expected with :content_file: [forms()](#mech-forms), [current_form()](#mech-current_form), [links()](#mech-links), [title()](#mech-title), [content(...)](#mech-content), [text()](#mech-text), all content-handling methods, all link methods, all image methods, all form methods, all field methods, [save_content(...)](#mech-save_content-filename-opts), [dump_links(...)](#mech-dump_links-fh-absolute), [dump_images(...)](#mech-dump_images-fh-absolute), [dump_forms(...)](#mech-dump_forms-fh), [dump_text(...)](#mech-dump_text-fh)

$mech->post( $uri, content => $content )

POSTs $content to $uri. Returns an HTTP::Response object. $uri can be a well-formed URI string, a URI object, or a WWW::Mechanize::Link object.

$mech->put( $uri, content => $content )

PUTs $content to $uri. Returns an HTTP::Response object. $uri can be a well-formed URI string, a URI object, or a WWW::Mechanize::Link object.

my $res = $mech->head( $uri );
my $res = $mech->head( $uri , $field_name => $value, ... );

$mech->head ($uri )

Performs a HEAD request to $uri. Returns an HTTP::Response object. $uri can be a well-formed URI string, a URI object, or a WWW::Mechanize::Link object.

$mech->reload()

Acts like the reload button in a browser: repeats the current request. The history (as per the back() method) is not altered.

Returns the HTTP::Response object from the reload, or undef if there's no current request.

$mech->back()

The equivalent of hitting the "back" button in a browser. Returns to the previous page. Won't go back past the first page. (Really, what would it do if it could?)

Returns true if it could go back, or false if not.

$mech->clear_history()

This deletes all the history entries and returns true.

$mech->history_count()

This returns the number of items in the browser history. This number does include the most recently made request.

$mech->history($n)

This returns the _n_th item in history. The 0th item is the most recent request and response, which would be acted on by methods like [find_link()](#mech-find_link). The 1st item is the state you'd return to if you called [back()](#mech-back).

The maximum useful value for $n is $mech->history_count - 1. Requests beyond that bound will return undef.

History items are returned as hash references, in the form:

{ req => $http_request, res => $http_response }

STATUS METHODS

$mech->success()

Returns a boolean telling whether the last request was successful. If there hasn't been an operation yet, returns false.

This is a convenience function that wraps $mech->res->is_success.

$mech->uri()

Returns the current URI as a URI object. This object stringifies to the URI itself.

$mech->response() / $mech->res()

Return the current response as an HTTP::Response object.

Synonym for $mech->response().

$mech->status()

Returns the HTTP status code of the response. This is a 3-digit number like 200 for OK, 404 for not found, and so on.

$mech->ct() / $mech->content_type()

Returns the content type of the response.

$mech->base()

Returns the base URI for the current response

$mech->forms()

When called in a list context, returns a list of the forms found in the last fetched page. In a scalar context, returns a reference to an array with those forms. The forms returned are all HTML::Form objects.

$mech->current_form()

Returns the current form as an HTML::Form object.

$mech->links()

When called in a list context, returns a list of the links found in the last fetched page. In a scalar context it returns a reference to an array with those links. Each link is a WWW::Mechanize::Link object.

$mech->is_html()

Returns true/false on whether our content is HTML, according to the HTTP headers.

$mech->title()

Returns the contents of the <TITLE> tag, as parsed by HTML::HeadParser. Returns undef if the content is not HTML.

$mech->redirects()

Convenience method to get the redirects from the most recent HTTP::Response.

Note that you can also use is_redirect to see if the most recent response was a redirect like this.

$mech->get($url);
do_stuff() if $mech->res->is_redirect;

CONTENT-HANDLING METHODS

$mech->content(...)

Returns the content that the mech uses internally for the last page fetched. Ordinarily this is the same as $mech->response()->decoded_content(), but this may differ for HTML documents if [update_html](#mech-update_html-html) is overloaded (in which case the value passed to the base-class implementation of same will be returned), and/or extra named arguments are passed to content():

$mech->content( format => 'text' )

Returns a text-only version of the page, with all HTML markup stripped. This feature requires HTML::TreeBuilder version 5 or higher to be installed, or a fatal error will be thrown. This works only if the contents are HTML.

$mech->content( base_href => [$base_href|undef] )

Returns the HTML document, modified to contain a <base href="$base_href"> mark-up in the header. $base_href is $mech->base() if not specified. This is handy to pass the HTML to e.g. HTML::Display. This works only if the contents are HTML.

$mech->content( raw => 1 )

Returns $self->response()->content(), i.e. the raw contents from the response.

$mech->content( decoded_by_headers => 1 )

Returns the content after applying all Content-Encoding headers but with not additional mangling.

$mech->content( charset => $charset )

Returns $self->response()->decoded_content(charset => $charset) (see HTTP::Response for details).

To preserve backwards compatibility, additional parameters will be ignored unless none of raw | decoded_by_headers | charset is specified and the text is HTML, in which case an error will be triggered.

A fresh instance of WWW::Mechanize will return undef when $mech->content() is called, because no content is present before a request has been made.

$mech->text()

Returns the text of the current HTML content. If the content isn't HTML, $mech will die.

The text is extracted by parsing the content, and then the extracted text is cached, so don't worry about performance of calling this repeatedly.

LINK METHODS

$mech->links()

Lists all the links on the current page. Each link is a WWW::Mechanize::Link object. In list context, returns a list of all links. In scalar context, returns an array reference of all links.

$mech->follow_link(...)

Follows a specified link on the page. You specify the match to be found using the same params that [find_link()](#mech-find_link) uses.

Here some examples:

3rd link called "download"

  $mech->follow_link( text => 'download', n => 3 );

first link where the URL has "download" in it, regardless of case:

  $mech->follow_link( url_regex => qr/download/i );

or

  $mech->follow_link( url_regex => qr/(?i:download)/ );

3rd link on the page

  $mech->follow_link( n => 3 );

the link with the url

  $mech->follow_link( url => '/other/page' );

or

  $mech->follow_link( url => 'http://example.com/page' );

Returns the result of the GET method (an HTTP::Response object) if a link was found.

If the page has no links, or the specified link couldn't be found, returns undef. If autocheck is enabled an exception will be thrown instead.

$mech->find_link( ... )

Finds a link in the currently fetched page. It returns a WWW::Mechanize::Link object which describes the link. (You'll probably be most interested in the [url()](https://metacpan.org/pod/WWW%3A%3AMechanize%3A%3ALink#link-url) property.) If it fails to find a link it returns undef.

You can take the URL part and pass it to the get() method. If that's your plan, you might as well use the [follow_link()](#mech-follow_link) method directly, since it does the get() for you automatically.

Note that <FRAME SRC="..."> tags are parsed out of the HTML and treated as links so this method works with them.

You can select which link to find by passing in one or more of these key/value pairs:

text => 'string', and text_regex => qr/regex/,

text matches the text of the link against string, which must be an exact match. To select a link with text that is exactly "download", use

  $mech->find_link( text => 'download' );

text_regex matches the text of the link against regex. To select a link with text that has "download" anywhere in it, regardless of case, use

  $mech->find_link( text_regex => qr/download/i );

Note that the text extracted from the page's links are trimmed. For example, <a> foo </a> is stored as 'foo', and searching for leading or trailing spaces will fail.

url => 'string', and url_regex => qr/regex/,

Matches the URL of the link against string or regex, as appropriate. The URL may be a relative URL, like foo/bar.html, depending on how it's coded on the page.

url_abs => string and url_abs_regex => regex

Matches the absolute URL of the link against string or regex, as appropriate. The URL will be an absolute URL, even if it's relative in the page.

name => string and name_regex => regex

Matches the name of the link against string or regex, as appropriate.

rel => string and rel_regex => regex

Matches the rel of the link against string or regex, as appropriate. This can be used to find stylesheets, favicons, or links the author of the page does not want bots to follow.

id => string and id_regex => regex

Matches the attribute 'id' of the link against string or regex, as appropriate.

class => string and class_regex => regex

Matches the attribute 'class' of the link against string or regex, as appropriate.

tag => string and tag_regex => regex

Matches the tag that the link came from against string or regex, as appropriate. The tag_regex is probably most useful to check for more than one tag, as in:

  $mech->find_link( tag_regex => qr/^(a|frame)$/ );

The tags and attributes looked at are defined below.

If n is not specified, it defaults to 1. Therefore, if you don't specify any params, this method defaults to finding the first link on the page.

Note that you can specify multiple text or URL parameters, which will be ANDed together. For example, to find the first link with text of "News" and with "cnn.com" in the URL, use:

$mech->find_link( text => 'News', url_regex => qr/cnn\.com/ );

The return value is a reference to an array containing a WWW::Mechanize::Link object for every link in [$self->content](#mech-content).

The links come from the following:

  • <a href=...>
  • <area href=...>
  • <frame src=...>
  • <iframe src=...>
  • <link href=...>
  • <meta content=...>

$mech->find_all_links( ... )

Returns all the links on the current page that match the criteria. The method for specifying link criteria is the same as in [find_link()](#mech-find_link). Each of the links returned is a WWW::Mechanize::Link object.

In list context, find_all_links() returns a list of the links. Otherwise, it returns a reference to the list of links.

find_all_links() with no parameters returns all links in the page.

$mech->find_all_inputs( ... criteria ... )

find_all_inputs() returns an array of all the input controls in the current form whose properties match all of the regexes passed in. The controls returned are all descended from HTML::Form::Input. See "INPUTS" in HTML::Form for details.

If no criteria are passed, all inputs will be returned.

If there is no current page, there is no form on the current page, or there are no submit controls in the current form then the return will be an empty array.

You may use a regex or a literal string:

# get all textarea controls whose names begin with "customer"
my @customer_text_inputs = $mech->find_all_inputs(
    type       => 'textarea',
    name_regex => qr/^customer/,
);

# get all text or textarea controls called "customer"
my @customer_text_inputs = $mech->find_all_inputs(
    type_regex => qr/^(text|textarea)$/,
    name       => 'customer',
);

$mech->find_all_submits( ... criteria ... )

find_all_submits() does the same thing as [find_all_inputs()](#mech-find_all_inputs-criteria) except that it only returns controls that are submit controls, ignoring other types of input controls like text and checkboxes.

IMAGE METHODS

$mech->images

Lists all the images on the current page. Each image is a WWW::Mechanize::Image object. In list context, returns a list of all images. In scalar context, returns an array reference of all images.

$mech->find_image()

Finds an image in the current page. It returns a WWW::Mechanize::Image object which describes the image. If it fails to find an image it returns undef.

You can select which image to find by passing in one or more of these key/value pairs:

alt => 'string' and alt_regex => qr/regex/

alt matches the ALT attribute of the image against string, which must be an exact match. To select a image with an ALT tag that is exactly "download", use

  $mech->find_image( alt => 'download' );

alt_regex matches the ALT attribute of the image against a regular expression. To select an image with an ALT attribute that has "download" anywhere in it, regardless of case, use

  $mech->find_image( alt_regex => qr/download/i );

url => 'string' and url_regex => qr/regex/

Matches the URL of the image against string or regex, as appropriate. The URL may be a relative URL, like foo/bar.html, depending on how it's coded on the page.

url_abs => string and url_abs_regex => regex

Matches the absolute URL of the image against string or regex, as appropriate. The URL will be an absolute URL, even if it's relative in the page.

tag => string and tag_regex => regex

Matches the tag that the image came from against string or regex, as appropriate. The tag_regex is probably most useful to check for more than one tag, as in:

  $mech->find_image( tag_regex => qr/^(img|input)$/ );

The tags supported are <img> and <input>.

id => string and id_regex => regex

id matches the id attribute of the image against string, which must be an exact match. To select an image with the exact id "download-image", use

  $mech->find_image( id => 'download-image' );

id_regex matches the id attribute of the image against a regular expression. To select the first image with an id that contains "download" anywhere in it, use

  $mech->find_image( id_regex => qr/download/ );

classs => string and class_regex => regex

class matches the class attribute of the image against string, which must be an exact match. To select an image with the exact class "img-fuid", use

  $mech->find_image( class => 'img-fluid' );

To select an image with the class attribute "rounded float-left", use

  $mech->find_image( class => 'rounded float-left' );

Note that the classes have to be matched as a complete string, in the exact order they appear in the website's source code.

class_regex matches the class attribute of the image against a regular expression. Use this if you want a partial class name, or if an image has several classes, but you only care about one.

To select the first image with the class "rounded", where there are multiple images that might also have either class "float-left" or "float-right", use

  $mech->find_image( class_regex => qr/\brounded\b/ );

Selecting an image with multiple classes where you do not care about the order they appear in the website's source code is not currently supported.

If n is not specified, it defaults to 1. Therefore, if you don't specify any params, this method defaults to finding the first image on the page.

Note that you can specify multiple ALT or URL parameters, which will be ANDed together. For example, to find the first image with ALT text of "News" and with "cnn.com" in the URL, use:

$mech->find_image( image => 'News', url_regex => qr/cnn\.com/ );

The return value is a reference to an array containing a WWW::Mechanize::Image object for every image in [$mech->content](#mech-content).

$mech->find_all_images( ... )

Returns all the images on the current page that match the criteria. The method for specifying image criteria is the same as in [find_image()](#mech-find_image). Each of the images returned is a WWW::Mechanize::Image object.

In list context, find_all_images() returns a list of the images. Otherwise, it returns a reference to the list of images.

find_all_images() with no parameters returns all images in the page.

FORM METHODS

These methods let you work with the forms on a page. The idea is to choose a form that you'll later work with using the field methods below.

$mech->forms

Lists all the forms on the current page. Each form is an HTML::Form object. In list context, returns a list of all forms. In scalar context, returns an array reference of all forms.

$mech->form_number($number)

Selects the _number_th form on the page as the target for subsequent calls to [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y). Also returns the form that was selected.

If it is found, the form is returned as an HTML::Form object and set internally for later use with Mech's form methods such as [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y). When called in a list context, the number of the found form is also returned as a second value.

Emits a warning and returns undef if no form is found.

The first form is number 1, not zero.

$mech->form_action( $action )

Selects a form by action, using a regex containing $action. If there is more than one form on the page matching that action, then the first one is used, and a warning is generated.

If it is found, the form is returned as an HTML::Form object and set internally for later use with Mech's form methods such as [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y).

Returns undef if no form is found.

$mech->form_name( $name [, \%args ] )

Selects a form by name.

By default, the first form that has this name will be returned.

my $form = $mech->form_name("order_form");

If you want the second, third or nth match, pass an optional arguments hash reference as the final parameter with a key n to pick which instance you want. The numbering starts at 1.

my $third_product_form = $mech->form_name("buy_now", { n => 3 });

If the n parameter is not passed, and there is more than one form on the page with that name, then the first one is used, and a warning is generated.

If it is found, the form is returned as an HTML::Form object and set internally for later use with Mech's form methods such as [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y).

Returns undef if no form is found.

$mech->form_id( $id [, \%args ] )

Selects a form by ID.

By default, the first form that has this ID will be returned.

my $form = $mech->form_id("order_form");

Although the HTML specification requires the ID to be unique within a page, some pages might not adhere to that. If you want the second, third or nth match, pass an optional arguments hash reference as the final parameter with a key n to pick which instance you want. The numbering starts at 1.

my $third_product_form = $mech->form_id("buy_now", { n => 3 });

If the n parameter is not passed, and there is more than one form on the page with that ID, then the first one is used, and a warning is generated.

If it is found, the form is returned as an HTML::Form object and set internally for later use with Mech's form methods such as [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y).

If no form is found it returns undef. This will also trigger a warning, unless quiet is enabled.

$mech->all_forms_with_fields( @fields )

Selects a form by passing in a list of field names it must contain. All matching forms (perhaps none) are returned as a list of HTML::Form objects.

$mech->form_with_fields( @fields, [ \%args ] )

Selects a form by passing in a list of field names it must contain. By default, the first form that matches all of these field names will be returned.

my $form = $mech->form_with_fields( qw/sku quantity add_to_cart/ );

If you want the second, third or nth match, pass an optional arguments hash reference as the final parameter with a key n to pick which instance you want. The numbering starts at 1.

my $form = $mech->form_with_fields( 'sky', 'qty', { n => 2 } );

If the n parameter is not passed, and there is more than one form on the page with that ID, then the first one is used, and a warning is generated.

If it is found, the form is returned as an HTML::Form object and set internally for later used with Mech's form methods such as [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y).

Returns undef and emits a warning if no form is found.

Note that this functionality requires libwww-perl 5.69 or higher.

$mech->all_forms_with( $attr1 => $value1, $attr2 => $value2, ... )

Searches for forms with arbitrary attribute/value pairs within the <form> tag. When given more than one pair, all criteria must match. Using undef as value means that the attribute in question must not be present.

All matching forms (perhaps none) are returned as a list of HTML::Form objects.

$mech->form_with( $attr1 => $value1, $attr2 => $value2, ..., [ \%args ] )

Searches for forms with arbitrary attribute/value pairs within the <form> tag. When given more than one pair, all criteria must match. Using undef as value means that the attribute in question must not be present.

By default, the first form that matches all criteria will be returned.

my $form = $mech->form_with( name => 'order_form', method => 'POST' );

If you want the second, third or nth match, pass an optional arguments hash reference as the final parameter with a key n to pick which instance you want. The numbering starts at 1.

my $form = $mech->form_with( method => 'POST', { n => 4 } );

If the n parameter is not passed, and there is more than one form on the page matching these criteria, then the first one is used, and a warning is generated.

If it is found, the form is returned as an HTML::Form object and set internally for later used with Mech's form methods such as [field()](#mech-field-name-value-number) and [click()](#mech-click-button-x-y).

Returns undef if no form is found.

FIELD METHODS

These methods allow you to set the values of fields in a given form.

$mech->field( $name, $value, $number )

$mech->field( $name, \@values, $number )

$mech->field( $name, \@file_upload_values, $number )

Given the name of a field, set its value to the value specified. This applies to the current form (as set by the [form_name()](#mech-form_name-name-args) or [form_number()](#mech-form_number-number) method or defaulting to the first form on the page).

If the field is of type "file", its value should be an arrayref. Example:

$mech->field( $file_input, ['/tmp/file.txt'] );

Value examples for "file" inputs, followed by explanation of what each index mean:

# 0: filepath      1: filename    3: headers
['/tmp/file.txt']
['/tmp/file.txt', 'filename.txt']
['/tmp/file.txt', 'filename.txt', @headers]
['/tmp/file.txt', 'filename.txt', Content => 'some content']
[undef,           'filename.txt', Content => 'content here']

Index 0 is the filepath that will be read from disk. Index 1 is the filename which will be used in the HTTP request body; if not given, filepath (index 0) is used instead. If <Content = 'content here'>> is used as shown, then filepath will be ignored.

The optional $number parameter is used to distinguish between two fields with the same name. The fields are numbered from 1.

$mech->select($name, $value)

$mech->select($name, \@values)

Given the name of a select field, set its value to the value specified. If the field is not <select multiple> and the $value is an array, only the first value will be set. [Note: the documentation previously claimed that only the last value would be set, but this was incorrect.] Passing $value as a hash with an n key selects an item by number (e.g. {n => 3} or {n => [2,4]}). The numbering starts at 1. This applies to the current form.

If you have a field with <select multiple> and you pass a single $value, then $value will be added to the list of fields selected, without clearing the others. However, if you pass an array reference, then all previously selected values will be cleared.

Returns true on successfully setting the value. On failure, returns false and calls $self->warn() with an error message.

$mech->set_fields( $name => $value ... )

$mech->set_fields( $name => \@value_and_instance_number )

$mech->set_fields( $name => \$value_instance_number )

$mech->set_fields( $name => \@file_upload )

This method sets multiple fields of the current form. It takes a list of field name and value pairs. If there is more than one field with the same name, the first one found is set. If you want to select which of the duplicate field to set, use a value which is an anonymous array which has the field value and its number as the 2 elements.

    # set the second $name field to 'foo'
    $mech->set_fields( $name => [ 'foo', 2 ] );

The value of a field of type "file" should be an arrayref as described in [field()](https://metacpan.org/pod/%24mech-%3Efield%28%20%24name%2C%20%24value%2C%20%24number%20%29). Examples:

    $mech->set_fields( $file_field => ['/tmp/file.txt'] );
    $mech->set_fields( $file_field => ['/tmp/file.txt', 'filename.txt'] );

The value for a "file" input can also be an arrayref containing an arrayref and a number, as documented in [submit_form()](https://metacpan.org/pod/%24mech-%3Esubmit_form%28%20...%20%29). The number will be used to find the field in the form. Example:

    $mech->set_fields( $file_field => [['/tmp/file.txt'], 1] );

The fields are numbered from 1.

For fields that have a predefined set of values, you may also provide a reference to an integer, if you don't know the options for the field, but you know you just want (e.g.) the first one.

    # select the first value in the $name select box
    $mech->set_fields( $name => \0 );
    # select the last value in the $name select box
    $mech->set_fields( $name => \-1 );

This applies to the current form.

$mech->set_visible( @criteria )

This method sets fields of the current form without having to know their names. So if you have a login screen that wants a username and password, you do not have to fetch the form and inspect the source (or use the mech-dump utility, installed with WWW::Mechanize) to see what the field names are; you can just say

$mech->set_visible( $username, $password );

and the first and second fields will be set accordingly. The method is called set_visible because it acts only on visible fields; hidden form inputs are not considered. The order of the fields is the order in which they appear in the HTML source which is nearly always the order anyone viewing the page would think they are in, but some creative work with tables could change that; caveat user.

Each element in @criteria is either a field value or a field specifier. A field value is a scalar. A field specifier allows you to specify the type of input field you want to set and is denoted with an arrayref containing two elements. So you could specify the first radio button with

$mech->set_visible( [ radio => 'KCRW' ] );

Field values and specifiers can be intermixed, hence

$mech->set_visible( 'fred', 'secret', [ option => 'Checking' ] );

would set the first two fields to "fred" and "secret", and the next OPTION menu field to "Checking".

The possible field specifier types are: "text", "password", "hidden", "textarea", "file", "image", "submit", "radio", "checkbox" and "option".

set_visible returns the number of values set.

$mech->tick( $name, $value [, $set] )

"Ticks" the first checkbox that has both the name and value associated with it on the current form. If there is no value to the input, just pass an empty string as the value. Dies if there is no named checkbox for the value given, if a value is given. Passing in a false value as the third optional argument will cause the checkbox to be unticked. The third value does not need to be set if you wish to merely tick the box.

$mech->tick('extra', 'cheese');
$mech->tick('extra', 'mushrooms');

$mech->tick('no_value', ''); # <input type="checkbox" name="no_value">

$mech->untick($name, $value)

Causes the checkbox to be unticked. Shorthand for tick($name,$value,undef)

$mech->value( $name [, $number] )

Given the name of a field, return its value. This applies to the current form.

The optional $number parameter is used to distinguish between two fields with the same name. The fields are numbered from 1.

If the field is of type file (file upload field), the value is always cleared to prevent remote sites from downloading your local files. To upload a file, specify its file name explicitly.

$mech->click( $button [, $x, $y] )

Has the effect of clicking a button on the current form. The first argument is the name of the button to be clicked. The second and third arguments (optional) allow you to specify the (x,y) coordinates of the click.

If there is only one button on the form, $mech->click() with no arguments simply clicks that one button.

Returns an HTTP::Response object.

$mech->click_button( ... )

Has the effect of clicking a button on the current form by specifying its attributes. The arguments are a list of key/value pairs. Only one of name, id, number, input or value must be specified in the keys.

Dies if no button is found.

name => name

Clicks the button named name in the current form.

id => id

Clicks the button with the id id in the current form.

number => n

Clicks the _n_th button with type submit in the current form. Numbering starts at 1.

value => value

Clicks the button with the value value in the current form.

input => $inputobject

Clicks on the button referenced by $inputobject, an instance of HTML::Form::SubmitInput obtained e.g. from

  $mech->current_form()->find_input( undef, 'submit' )

$inputobject must belong to the current form.

x => x

y => y

These arguments (optional) allow you to specify the (x,y) coordinates of the click.

$mech->submit()

Submits the current form, without specifying a button to click. Actually, no button is clicked at all.

Returns an HTTP::Response object.

This used to be a synonym for $mech->click( 'submit' ), but is no longer so.

$mech->submit_form( ... )

This method lets you select a form from the previously fetched page, fill in its fields, and submit it. It combines the form_number/form_name, set_fields and click methods into one higher level call. Its arguments are a list of key/value pairs, all of which are optional.

fields => \%fields

Specifies the fields to be filled in the current form.

with_fields => \%fields

Probably all you need for the common case. It combines a smart form selector and data setting in one operation. It selects the first form that contains all fields mentioned in \%fields. This is nice because you don't need to know the name or number of the form to do this.

(calls [form_with_fields()](#mech-form_with_fields-fields-args) and [set_fields()](#mech-set_fields-name-value)).

If you choose with_fields, the fields option will be ignored. The form_number, form_name and form_id options will still be used. An exception will be thrown unless exactly one form matches all of the provided criteria.

form_number => n

Selects the _n_th form (calls [form_number()](#mech-form_number-number). If this param is not specified, the currently-selected form is used.

form_name => name

Selects the form named name (calls [form_name()](#mech-form_name-name-args))

form_id => ID

Selects the form with ID ID (calls [form_id()](#mech-form_name-name-args))

button => button

Clicks on button button (calls [click()](#mech-click-button-x-y))

x => x, y => y

Sets the x or y values for [click()](#mech-click-button-x-y)

strict_forms => bool

Sets the HTML::Form strict flag which causes form submission to croak if any of the passed fields don't exist on the page, and/or a value doesn't exist in a select element. By default HTML::Form sets this value to false.

This behavior can also be turned on globally by passing strict_forms => 1 to WWW::Mechanize->new. If you do that, you can still disable it for individual calls by passing strict_forms => 0 here.

If no form is selected, the first form found is used.

If button is not passed, then the [submit()](#mech-submit) method is used instead.

If you want to submit a file and get its content from a scalar rather than a file in the filesystem, you can use:

$mech->submit_form(with_fields => { logfile => [ [ undef, 'whatever', Content => $content ], 1 ] } );

Returns an HTTP::Response object.

MISCELLANEOUS METHODS

$mech->add_header( name => $value [, name => $value... ] )

Sets HTTP headers for the agent to add or remove from the HTTP request.

$mech->add_header( Encoding => 'text/klingon' );

If a value is undef, then that header will be removed from any future requests. For example, to never send a Referer header:

$mech->add_header( Referer => undef );

If you want to delete a header, use delete_header.

Returns the number of name/value pairs added.

NOTE: This method was very different in WWW::Mechanize before 1.00. Back then, the headers were stored in a package hash, not as a member of the object instance. Calling add_header() would modify the headers for every WWW::Mechanize object, even after your object no longer existed.

$mech->delete_header( name [, name ... ] )

Removes HTTP headers from the agent's list of special headers. For instance, you might need to do something like:

# Don't send a Referer for this URL
$mech->add_header( Referer => undef );

# Get the URL
$mech->get( $url );

# Back to the default behavior
$mech->delete_header( 'Referer' );

$mech->quiet(true/false)

Allows you to suppress warnings to the screen.

$mech->quiet(0); # turns on warnings (the default)
$mech->quiet(1); # turns off warnings
$mech->quiet();  # returns the current quietness status

$mech->autocheck(true/false)

Allows you to enable and disable autochecking.

Autocheck checks each request made to see if it was successful. This saves you the trouble of manually checking yourself. Any errors found are errors, not warnings. Please see [new](#new) for more details.

$mech->autocheck(1); # turns on automatic request checking (the default)
$mech->autocheck(0); # turns off automatic request checking
$mech->autocheck();  # returns the current autocheck status

$mech->stack_depth( $max_depth )

Get or set the page stack depth. Use this if you're doing a lot of page scraping and running out of memory.

A value of 0 means "no history at all." By default, the max stack depth is humongously large, effectively keeping all history.

$mech->save_content( $filename, %opts )

Dumps the contents of [$mech->content](#mech-content) into $filename. $filename will be overwritten. Dies if there are any errors.

If the content type does not begin with "text/", then the content is saved in binary mode (i.e. binmode() is set on the output filehandle).

Additional arguments can be passed as key/value pairs:

$mech->save_content( $filename, binary => 1 )

Filehandle is set with binmode to :raw and contents are taken calling $self->content(decoded_by_headers => 1). Same as calling:

  $mech->save_content( $filename, binmode => ':raw',
                       decoded_by_headers => 1 );

This should be the safest way to save contents verbatim.

$mech->save_content( $filename, binmode => $binmode )

Filehandle is set to binary mode. If $binmode begins with ':', it is passed as a parameter to binmode:

  binmode $fh, $binmode;

otherwise the filehandle is set to binary mode if $binmode is true:

  binmode $fh;

all other arguments

are passed as-is to [$mech->content(%opts)](#mech-content). In particular, decoded_by_headers might come handy if you want to revert the effect of line compression performed by the web server but without further interpreting the contents (e.g. decoding it according to the charset).

$mech->dump_headers( [$fh] )

Prints a dump of the HTTP response headers for the most recent response. If $fh is not specified or is undef, it dumps to STDOUT.

Unlike the rest of the dump_* methods, $fh can be a scalar. It will be used as a file name.

$mech->dump_links( [[$fh], $absolute] )

Prints a dump of the links on the current page to $fh. If $fh is not specified or is undef, it dumps to STDOUT.

If $absolute is true, links displayed are absolute, not relative.

$mech->dump_images( [[$fh], $absolute] )

Prints a dump of the images on the current page to $fh. If $fh is not specified or is undef, it dumps to STDOUT.

If $absolute is true, links displayed are absolute, not relative.

The output will include empty lines for images that have no src attribute and therefore no URL.

$mech->dump_forms( [$fh] )

Prints a dump of the forms on the current page to $fh. If $fh is not specified or is undef, it dumps to STDOUT. Running the following:

my $mech = WWW::Mechanize->new();
$mech->get("https://www.google.com/");
$mech->dump_forms;

will print:

GET https://www.google.com/search [f]
  ie=ISO-8859-1                  (hidden readonly)
  hl=en                          (hidden readonly)
  source=hp                      (hidden readonly)
  biw=                           (hidden readonly)
  bih=                           (hidden readonly)
  q=                             (text)
  btnG=Google Search             (submit)
  btnI=I'm Feeling Lucky         (submit)
  gbv=1                          (hidden readonly)

$mech->dump_text( [$fh] )

Prints a dump of the text on the current page to $fh. If $fh is not specified or is undef, it dumps to STDOUT.

OVERRIDDEN LWP::UserAgent METHODS

$mech->clone()

Clone the mech object. The clone will be using the same cookie jar as the original mech.

$mech->redirect_ok()

An overloaded version of redirect_ok() in LWP::UserAgent. This method is used to determine whether a redirection in the request should be followed.

Note that WWW::Mechanize's constructor pushes POST on to the agent's requests_redirectable list.

$mech->request( $request [, $arg [, $size]])

Overloaded version of request() in LWP::UserAgent. Performs the actual request. Normally, if you're using WWW::Mechanize, it's because you don't want to deal with this level of stuff anyway.

Note that $request will be modified.

Returns an HTTP::Response object.

$mech->update_html( $html )

Allows you to replace the HTML that the mech has found. Updates the forms and links parse-trees that the mech uses internally.

Say you have a page that you know has malformed output, and you want to update it so the links come out correctly:

my $html = $mech->content;
$html =~ s[</option>.{0,3}</td>][</option></select></td>]isg;
$mech->update_html( $html );

This method is also used internally by the mech itself to update its own HTML content when loading a page. This means that if you would like to systematically perform the above HTML substitution, you would overload update_html in a subclass thusly:

package MyMech;
use base 'WWW::Mechanize';

sub update_html {
    my ($self, $html) = @_;
    $html =~ s[</option>.{0,3}</td>][</option></select></td>]isg;
    $self->WWW::Mechanize::update_html( $html );
}

If you do this, then the mech will use the tidied-up HTML instead of the original both when parsing for its own needs, and for returning to you through [content()](#mech-content).

Overloading this method is also the recommended way of implementing extra validation steps (e.g. link checkers) for every HTML page received. [warn](#warn-messages) and [warn](#warn-messages) would then come in handy to signal validation errors.

$mech->credentials( $username, $password )

Provide credentials to be used for HTTP Basic authentication for all sites and realms until further notice.

The four argument form described in LWP::UserAgent is still supported.

$mech->get_basic_credentials( $realm, $uri, $isproxy )

Returns the credentials for the realm and URI.

$mech->clear_credentials()

Remove any credentials set up with credentials().

INHERITED UNCHANGED LWP::UserAgent METHODS

As a subclass of LWP::UserAgent, WWW::Mechanize inherits all of LWP::UserAgent's methods. Many of which are overridden or extended. The following methods are inherited unchanged. View the LWP::UserAgent documentation for their implementation descriptions.

This is not meant to be an inclusive list. LWP::UA may have added others.

$mech->head()

Inherited from LWP::UserAgent.

$mech->mirror()

Inherited from LWP::UserAgent.

$mech->simple_request()

Inherited from LWP::UserAgent.

$mech->is_protocol_supported()

Inherited from LWP::UserAgent.

$mech->prepare_request()

Inherited from LWP::UserAgent.

$mech->progress()

Inherited from LWP::UserAgent.

INTERNAL-ONLY METHODS

These methods are only used internally. You probably don't need to know about them.

$mech->_update_page($request, $response)

Updates all internal variables in $mech as if $request was just performed, and returns $response. The page stack is not altered by this method, it is up to caller (e.g. [request](#mech-request-request-arg-size)) to do that.

$mech->_modify_request( $req )

Modifies a HTTP::Request before the request is sent out, for both GET and POST requests.

We add a Referer header, as well as header to note that we can accept gzip encoded content, if Compress::Zlib is installed.

$mech->_make_request()

Convenience method to make it easier for subclasses like WWW::Mechanize::Cached to intercept the request.

$mech->_reset_page()

Resets the internal fields that track page parsed stuff.

$mech->_extract_links()

Extracts links from the content of a webpage, and populates the {links} property with WWW::Mechanize::Link objects.

$mech->_push_page_stack()

The agent keeps a stack of visited pages, which it can pop when it needs to go BACK and so on.

The current page needs to be pushed onto the stack before we get a new page, and the stack needs to be popped when BACK occurs.

Neither of these take any arguments, they just operate on the $mech object.

warn( @messages )

Centralized warning method, for diagnostics and non-fatal problems. Defaults to calling CORE::warn, but may be overridden by setting onwarn in the constructor.

die( @messages )

Centralized error method. Defaults to calling CORE::die, but may be overridden by setting onerror in the constructor.

BEST PRACTICES

The default settings can get you up and running quickly, but there are settings you can change in order to make your life easier.

autocheck

autocheck can save you the overhead of checking status codes for success. You may outgrow it as your needs get more sophisticated, but it's a safe option to start with.

  my $agent = WWW::Mechanize->new( autocheck => 1 );

cookie_jar

You are encouraged to install Mozilla::PublicSuffix and use HTTP::CookieJar::LWP as your cookie jar. HTTP::CookieJar::LWP provides a better security model matching that of current Web browsers when Mozilla::PublicSuffix is installed.

  use HTTP::CookieJar::LWP ();

  my $jar = HTTP::CookieJar::LWP->new;
  my $agent = WWW::Mechanize->new( cookie_jar => $jar );

protocols_allowed

This option is inherited directly from LWP::UserAgent. It may be used to allow arbitrary protocols.

  my $agent = WWW::Mechanize->new(
      protocols_allowed => [ 'http', 'https' ]
  );

This will prevent you from inadvertently following URLs like file:///etc/passwd

protocols_forbidden

This option is also inherited directly from LWP::UserAgent. It may be used to deny arbitrary protocols.

  my $agent = WWW::Mechanize->new(
      protocols_forbidden => [ 'file', 'mailto', 'ssh', ]
  );

This will prevent you from inadvertently following URLs like file:///etc/passwd

strict_forms

Consider turning on the strict_forms option when you create a new Mech. This will perform a helpful sanity check on form fields every time you are submitting a form, which can save you a lot of debugging time.

  my $agent = WWW::Mechanize->new( strict_forms => 1 );

If you do not want to have this option globally, you can still turn it on for individual forms.

  $agent->submit_form( fields => { foo => 'bar' } , strict_forms => 1 );

WWW::MECHANIZE'S GIT REPOSITORY

WWW::Mechanize is hosted at GitHub.

Repository: https://github.com/libwww-perl/WWW-Mechanize. Bugs: https://github.com/libwww-perl/WWW-Mechanize/issues.

OTHER DOCUMENTATION

Spidering Hacks, by Kevin Hemenway and Tara Calishain

Spidering Hacks from O'Reilly (http://www.oreilly.com/catalog/spiderhks/) is a great book for anyone wanting to know more about screen-scraping and spidering.

There are six hacks that use Mech or a Mech derivative:

  • #21 WWW::Mechanize 101
  • #22 Scraping with WWW::Mechanize
  • #36 Downloading Images from Webshots
  • #44 Archiving Yahoo! Groups Messages with WWW::Yahoo::Groups
  • #64 Super Author Searching
  • #73 Scraping TV Listings

The book was also positively reviewed on Slashdot: http://books.slashdot.org/article.pl?sid=03/12/11/2126256

ONLINE RESOURCES AND SUPPORT

WWW::Mechanize mailing list

The Mech mailing list is at http://groups.google.com/group/www-mechanize-users and is specific to Mechanize, unlike the LWP mailing list below. Although it is a users list, all development discussion takes place here, too.

LWP mailing list

The LWP mailing list is at http://lists.perl.org/showlist.cgi?name=libwww, and is more user-oriented and well-populated than the WWW::Mechanize list.

Perlmonks

http://perlmonks.org is an excellent community of support, and many questions about Mech have already been answered there.

WWW::Mechanize::Examples

A random array of examples submitted by users, included with the Mechanize distribution.

ARTICLES ABOUT WWW::MECHANIZE

http://www.ibm.com/developerworks/linux/library/wa-perlsecure/

IBM article "Secure Web site access with Perl"

http://www.oreilly.com/catalog/googlehks2/chapter/hack84.pdf

Leland Johnson's hack #84 in Google Hacks, 2nd Edition is an example of a production script that uses WWW::Mechanize and HTML::TableContentParser. It takes in keywords and returns the estimated price of these keywords on Google's AdWords program.

http://www.perl.com/pub/a/2004/06/04/recorder.html

Linda Julien writes about using HTTP::Recorder to create WWW::Mechanize scripts.

http://www.developer.com/lang/other/article.php/3454041

Jason Gilmore's article on using WWW::Mechanize for scraping sales information from Amazon and eBay.

http://www.perl.com/pub/a/2003/01/22/mechanize.html

Chris Ball's article about using WWW::Mechanize for scraping TV listings.

http://www.stonehenge.com/merlyn/LinuxMag/col47.html

Randal Schwartz's article on scraping Yahoo News for images. It's already out of date: He manually walks the list of links hunting for matches, which wouldn't have been necessary if the [find_link()](#mech-find_link) method existed at press time.

http://www.perladvent.org/2002/16th/

WWW::Mechanize on the Perl Advent Calendar, by Mark Fowler.

http://www.linux-magazin.de/ausgaben/2004/03/datenruessel/

Michael Schilli's article on Mech and WWW::Mechanize::Shell for the German magazine Linux Magazin.

Other modules that use Mechanize

Here are modules that use or subclass Mechanize. Let me know of any others:

Finance::Bank::LloydsTSB

HTTP::Recorder

Acts as a proxy for web interaction, and then generates WWW::Mechanize scripts.

Win32::IE::Mechanize

Just like Mech, but using Microsoft Internet Explorer to do the work.

WWW::Bugzilla

WWW::Google::Groups

WWW::Hotmail

WWW::Mechanize::Cached

WWW::Mechanize::Cached::GZip

WWW::Mechanize::FormFiller

WWW::Mechanize::Shell

WWW::Mechanize::Sleepy

WWW::Mechanize::SpamCop

WWW::Mechanize::Timed

WWW::SourceForge

WWW::Yahoo::Groups

WWW::Scripter

ACKNOWLEDGEMENTS

Thanks to the numerous people who have helped out on WWW::Mechanize in one way or another, including Kirrily Robert for the original WWW::Automate, Lyle Hopkins, Damien Clark, Ansgar Burchardt, Gisle Aas, Jeremy Ary, Hilary Holz, Rafael Kitover, Norbert Buchmuller, Dave Page, David Sainty, H.Merijn Brand, Matt Lawrence, Michael Schwern, Adriano Ferreira, Miyagawa, Peteris Krumins, Rafael Kitover, David Steinbrunner, Kevin Falcone, Mike O'Regan, Mark Stosberg, Uri Guttman, Peter Scott, Philippe Bruhat, Ian Langworth, John Beppu, Gavin Estey, Jim Brandt, Ask Bjoern Hansen, Greg Davies, Ed Silva, Mark-Jason Dominus, Autrijus Tang, Mark Fowler, Stuart Children, Max Maischein, Meng Wong, Prakash Kailasa, Abigail, Jan Pazdziora, Dominique Quatravaux, Scott Lanning, Rob Casey, Leland Johnson, Joshua Gatcomb, Julien Beasley, Abe Timmerman, Peter Stevens, Pete Krawczyk, Tad McClellan, and the late great Iain Truskett.

AUTHOR

Andy Lester <andy at petdance.com>

COPYRIGHT AND LICENSE

This software is copyright (c) 2004 by Andy Lester.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.


Download Details:

Author: libwww-perl
Source Code: https://github.com/libwww-perl/WWW-Mechanize

License: View license

#perl 

What is GEEK

Buddha Community

WWW Mechanize: Handy Web Browsing in A Perl Object
Sasha  Lee

Sasha Lee

1650636000

Dl4clj: Clojure Wrapper for Deeplearning4j.

dl4clj

Port of deeplearning4j to clojure

Contact info

If you have any questions,

  • my email is will@yetanalytics.com
  • I'm will_hoyt in the clojurians slack
  • twitter is @FeLungz (don't check very often)

TODO

  • update examples dir
  • finish README
    • add in examples using Transfer Learning
  • finish tests
    • eval is missing regression tests, roc tests
    • nn-test is missing regression tests
    • spark tests need to be redone
    • need dl4clj.core tests
  • revist spark for updates
  • write specs for user facing functions
    • this is very important, match isnt strict for maps
    • provides 100% certianty of the input -> output flow
    • check the args as they come in, dispatch once I know its safe, test the pure output
  • collapse overlapping api namespaces
  • add to core use case flows

Features

Stable Features with tests

  • Neural Networks DSL
  • Early Stopping Training
  • Transfer Learning
  • Evaluation
  • Data import

Features being worked on for 0.1.0

  • Clustering (testing in progress)
  • Spark (currently being refactored)
  • Front End (maybe current release, maybe future release. Not sure yet)
  • Version of dl4j is 0.0.8 in this project. Current dl4j version is 0.0.9
  • Parallelism
  • Kafka support
  • Other items mentioned in TODO

Features being worked on for future releases

  • NLP
  • Computational Graphs
  • Reinforement Learning
  • Arbiter

Artifacts

NOT YET RELEASED TO CLOJARS

  • fork or clone to try it out

If using Maven add the following repository definition to your pom.xml:

<repository>
  <id>clojars.org</id>
  <url>http://clojars.org/repo</url>
</repository>

Latest release

With Leiningen:

n/a

With Maven:

n/a

<dependency>
  <groupId>_</groupId>
  <artifactId>_</artifactId>
  <version>_</version>
</dependency>

Usage

Things you need to know

All functions for creating dl4j objects return code by default

  • All of these functions have an option to return the dl4j object
    • :as-code? = false
  • This because all builders require the code representation of dl4j objects
    • this requirement is not going to change
  • INDarray creation fns default to objects, this is for convenience
    • :as-code? is still respected

API functions return code when all args are provided as code

API functions return the value of calling the wrapped method when args are provided as a mixture of objects and code or just objects

The tests are there to help clarify behavior, if you are unsure of how to use a fn, search the tests

  • for questions about spark, refer to the spark section bellow

Example of obj/code duality

(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]))

;; as code (the default)

(l/dense-layer-builder
 :activation-fn :relu
 :learning-rate 0.006
 :weight-init :xavier
 :layer-name "example layer"
 :n-in 10
 :n-out 1)

;; =>

(doto
 (org.deeplearning4j.nn.conf.layers.DenseLayer$Builder.)
 (.nOut 1)
 (.activation (dl4clj.constants/value-of {:activation-fn :relu}))
 (.weightInit (dl4clj.constants/value-of {:weight-init :xavier}))
 (.nIn 10)
 (.name "example layer")
 (.learningRate 0.006))

;; as an object

(l/dense-layer-builder
 :activation-fn :relu
 :learning-rate 0.006
 :weight-init :xavier
 :layer-name "example layer"
 :n-in 10
 :n-out 1
 :as-code? false)

;; =>

#object[org.deeplearning4j.nn.conf.layers.DenseLayer 0x69d7d160 "DenseLayer(super=FeedForwardLayer(super=Layer(layerName=example layer, activationFn=relu, weightInit=XAVIER, biasInit=NaN, dist=null, learningRate=0.006, biasLearningRate=NaN, learningRateSchedule=null, momentum=NaN, momentumSchedule=null, l1=NaN, l2=NaN, l1Bias=NaN, l2Bias=NaN, dropOut=NaN, updater=null, rho=NaN, epsilon=NaN, rmsDecay=NaN, adamMeanDecay=NaN, adamVarDecay=NaN, gradientNormalization=null, gradientNormalizationThreshold=NaN), nIn=10, nOut=1))"]

General usage examples

Importing data

Loading data from a file (here its a csv)


(ns my.ns
 (:require [dl4clj.datasets.input-splits :as s]
           [dl4clj.datasets.record-readers :as rr]
           [dl4clj.datasets.api.record-readers :refer :all]
           [dl4clj.datasets.iterators :as ds-iter]
           [dl4clj.datasets.api.iterators :refer :all]
           [dl4clj.helpers :refer [data-from-iter]]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; file splits (convert the data to records)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def poker-path "resources/poker-hand-training.csv")
;; this is not a complete dataset, it is just here to sever as an example

(def file-split (s/new-filesplit :path poker-path))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; record readers, (read the records created by the file split)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def csv-rr (initialize-rr! :rr (rr/new-csv-record-reader :skip-n-lines 0 :delimiter ",")
                                 :input-split file-split))

;; lets look at some data
(println (next-record! :rr csv-rr :as-code? false))
;; => #object[java.util.ArrayList 0x2473e02d [1, 10, 1, 11, 1, 13, 1, 12, 1, 1, 9]]
;; this is our first line from the csv


;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; record readers dataset iterators (turn our writables into a dataset)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def rr-ds-iter (ds-iter/new-record-reader-dataset-iterator
                 :record-reader csv-rr
                 :batch-size 1
                 :label-idx 10
                 :n-possible-labels 10))

;; we use our record reader created above
;; we want to see one example per dataset obj returned (:batch-size = 1)
;; we know our label is at the last index, so :label-idx = 10
;; there are 10 possible types of poker hands so :n-possible-labels = 10
;; you can also set :label-idx to -1 to use the last index no matter the size of the seq

(def other-rr-ds-iter (ds-iter/new-record-reader-dataset-iterator
                       :record-reader csv-rr
                       :batch-size 1
                       :label-idx -1
                       :n-possible-labels 10))

(str (next-example! :iter rr-ds-iter :as-code? false))
;; =>
;;===========INPUT===================
;;[1.00, 10.00, 1.00, 11.00, 1.00, 13.00, 1.00, 12.00, 1.00, 1.00]
;;=================OUTPUT==================
;;[0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 1.00]


;; and to show that :label-idx = -1 gives us the same output

(= (next-example! :iter rr-ds-iter :as-code? false)
   (next-example! :iter other-rr-ds-iter :as-code? false)) ;; => true

INDArrays and Datasets from clojure data structures


(ns my.ns
  (:require [nd4clj.linalg.factory.nd4j :refer [vec->indarray matrix->indarray
                                                indarray-of-zeros indarray-of-ones
                                                indarray-of-rand vec-or-matrix->indarray]]
            [dl4clj.datasets.new-datasets :refer [new-ds]]
            [dl4clj.datasets.api.datasets :refer [as-list]]
            [dl4clj.datasets.iterators :refer [new-existing-dataset-iterator]]
            [dl4clj.datasets.api.iterators :refer :all]
            [dl4clj.datasets.pre-processors :as ds-pp]
            [dl4clj.datasets.api.pre-processors :refer :all]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; INDArray creation
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;;TODO: consider defaulting to code

;; can create from a vector

(vec->indarray [1 2 3 4])
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x269df212 [1.00, 2.00, 3.00, 4.00]]

;; or from a matrix

(matrix->indarray [[1 2 3 4] [2 4 6 8]])
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x20aa7fe1
;; [[1.00, 2.00, 3.00, 4.00], [2.00, 4.00, 6.00, 8.00]]]


;; will fill in spareness with zeros

(matrix->indarray [[1 2 3 4] [2 4 6 8] [10 12]])
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x8b7796c
;;[[1.00, 2.00, 3.00, 4.00],
;; [2.00, 4.00, 6.00, 8.00],
;; [10.00, 12.00, 0.00, 0.00]]]

;; can create an indarray of all zeros with specified shape
;; defaults to :rows = 1 :columns = 1

(indarray-of-zeros :rows 3 :columns 2)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x6f586a7e
;;[[0.00, 0.00],
;; [0.00, 0.00],
;; [0.00, 0.00]]]

(indarray-of-zeros) ;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0xe59ffec 0.00]

;; and if only one is supplied, will get a vector of specified length

(indarray-of-zeros :rows 2)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x2899d974 [0.00, 0.00]]

(indarray-of-zeros :columns 2)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0xa5b9782 [0.00, 0.00]]

;; same considerations/defaults for indarray-of-ones and indarray-of-rand

(indarray-of-ones :rows 2 :columns 3)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x54f08662 [[1.00, 1.00, 1.00], [1.00, 1.00, 1.00]]]

(indarray-of-rand :rows 2 :columns 3)
;; all values are greater than 0 but less than 1
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x2f20293b [[0.85, 0.86, 0.13], [0.94, 0.04, 0.36]]]



;; vec-or-matrix->indarray is built into all functions which require INDArrays
;; so that you can use clojure data structures
;; but you still have the option of passing existing INDArrays

(def example-array (vec-or-matrix->indarray [1 2 3 4]))
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x5c44c71f [1.00, 2.00, 3.00, 4.00]]

(vec-or-matrix->indarray example-array)
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x607b03b0 [1.00, 2.00, 3.00, 4.00]]

(vec-or-matrix->indarray (indarray-of-rand :rows 2))
;; => #object[org.nd4j.linalg.cpu.nativecpu.NDArray 0x49143b08 [0.76, 0.92]]

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; data-set creation
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def ds-with-single-example (new-ds :input [1 2 3 4]
                                    :output [0.0 1.0 0.0]))

(as-list :ds ds-with-single-example :as-code? false)
;; =>
;; #object[java.util.ArrayList 0x5d703d12
;;[===========INPUT===================
;;[1.00, 2.00, 3.00, 4.00]
;;=================OUTPUT==================
;;[0.00, 1.00, 0.00]]]

(def ds-with-multiple-examples (new-ds
                                :input [[1 2 3 4] [2 4 6 8]]
                                :output [[0.0 1.0 0.0] [0.0 0.0 1.0]]))

(as-list :ds ds-with-multiple-examples :as-code? false)
;; =>
;;#object[java.util.ArrayList 0x29c7a9e2
;;[===========INPUT===================
;;[1.00, 2.00, 3.00, 4.00]
;;=================OUTPUT==================
;;[0.00, 1.00, 0.00],
;;===========INPUT===================
;;[2.00, 4.00, 6.00, 8.00]
;;=================OUTPUT==================
;;[0.00, 0.00, 1.00]]]

;; we can create a dataset iterator from the code which creates datasets
;; and set the labels for our outputs (optional)

(def ds-with-multiple-examples
  (new-ds
   :input [[1 2 3 4] [2 4 6 8]]
   :output [[0.0 1.0 0.0] [0.0 0.0 1.0]]))

;; iterator
(def training-rr-ds-iter
  (new-existing-dataset-iterator
   :dataset ds-with-multiple-examples
   :labels ["foo" "baz" "foobaz"]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; data-set normalization
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; this gathers statistics on the dataset and normalizes the data
;; and applies the transformation to all dataset objects in the iterator
(def train-iter-normalized
  (c/normalize-iter! :iter training-rr-ds-iter
                     :normalizer (ds-pp/new-standardize-normalization-ds-preprocessor)
                     :as-code? false))

;; above returns the normalized iterator
;; to get fit normalizer

(def the-normalizer
  (get-pre-processor train-iter-normalized))

Model configuration

Creating a neural network configuration with singe and multiple layers

(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.nn.conf.distributions :as dist]
            [dl4clj.nn.conf.input-pre-processor :as pp]
            [dl4clj.nn.conf.step-fns :as s-fn]))

;; nn/builder has 3 types of args
;; 1) args which set network configuration params
;; 2) args which set default values for layers
;; 3) args which set multi layer network configuration params

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; single layer nn configuration
;; here we are setting network configuration
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(nn/builder :optimization-algo :stochastic-gradient-descent
            :seed 123
            :iterations 1
            :minimize? true
            :use-drop-connect? false
            :lr-score-based-decay-rate 0.002
            :regularization? false
            :step-fn :default-step-fn
            :layers {:dense-layer {:activation-fn :relu
                                   :updater :adam
                                   :adam-mean-decay 0.2
                                   :adam-var-decay 0.1
                                   :learning-rate 0.006
                                   :weight-init :xavier
                                   :layer-name "single layer model example"
                                   :n-in 10
                                   :n-out 20}})

;; there are several options within a nn-conf map which can be configuration maps
;; or calls to fns
;; It doesn't matter which option you choose and you don't have to stay consistent
;; the list of params which can be passed as config maps or fn calls will
;; be enumerated at a later date

(nn/builder :optimization-algo :stochastic-gradient-descent
            :seed 123
            :iterations 1
            :minimize? true
            :use-drop-connect? false
            :lr-score-based-decay-rate 0.002
            :regularization? false
            :step-fn (s-fn/new-default-step-fn)
            :build? true
            ;; dont need to specify layer order, theres only one
            :layers (l/dense-layer-builder
                    :activation-fn :relu
                    :updater :adam
                    :adam-mean-decay 0.2
                    :adam-var-decay 0.1
                    :dist (dist/new-normal-distribution :mean 0 :std 1)
                    :learning-rate 0.006
                    :weight-init :xavier
                    :layer-name "single layer model example"
                    :n-in 10
                    :n-out 20))

;; these configurations are the same

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; multi-layer configuration
;; here we are also setting layer defaults
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; defaults will apply to layers which do not specify those value in their config

(nn/builder
 :optimization-algo :stochastic-gradient-descent
 :seed 123
 :iterations 1
 :minimize? true
 :use-drop-connect? false
 :lr-score-based-decay-rate 0.002
 :regularization? false
 :default-activation-fn :sigmoid
 :default-weight-init :uniform

 ;; we need to specify the layer order
 :layers {0 (l/activation-layer-builder
             :activation-fn :relu
             :updater :adam
             :adam-mean-decay 0.2
             :adam-var-decay 0.1
             :learning-rate 0.006
             :weight-init :xavier
             :layer-name "example first layer"
             :n-in 10
             :n-out 20)
          1 {:output-layer {:n-in 20
                            :n-out 2
                            :loss-fn :mse
                            :layer-name "example output layer"}}})

;; specifying multi-layer config params

(nn/builder
 ;; network args
 :optimization-algo :stochastic-gradient-descent
 :seed 123
 :iterations 1
 :minimize? true
 :use-drop-connect? false
 :lr-score-based-decay-rate 0.002
 :regularization? false

 ;; layer defaults
 :default-activation-fn :sigmoid
 :default-weight-init :uniform

 ;; the layers
 :layers {0 (l/activation-layer-builder
             :activation-fn :relu
             :updater :adam
             :adam-mean-decay 0.2
             :adam-var-decay 0.1
             :learning-rate 0.006
             :weight-init :xavier
             :layer-name "example first layer"
             :n-in 10
             :n-out 20)
          1 {:output-layer {:n-in 20
                            :n-out 2
                            :loss-fn :mse
                            :layer-name "example output layer"}}}
 ;; multi layer network args
 :backprop? true
 :backprop-type :standard
 :pretrain? false
 :input-pre-processors {0 (pp/new-zero-mean-pre-pre-processor)
                        1 {:unit-variance-processor {}}})

Configuration to Trained models

Multi Layer models

(ns my.ns
  (:require [dl4clj.datasets.iterators :as iter]
            [dl4clj.datasets.input-splits :as split]
            [dl4clj.datasets.record-readers :as rr]
            [dl4clj.optimize.listeners :as listener]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.nn.multilayer.multi-layer-network :as mln]
            [dl4clj.nn.api.model :refer [init! set-listeners!]]
            [dl4clj.nn.api.multi-layer-network :refer [evaluate-classification]]
            [dl4clj.datasets.api.record-readers :refer [initialize-rr!]]
            [dl4clj.eval.api.eval :refer [get-stats get-accuracy]]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; nn-conf -> multi-layer-network
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def nn-conf
  (nn/builder
   ;; network args
   :optimization-algo :stochastic-gradient-descent
   :seed 123 :iterations 1 :regularization? true

   ;; setting layer defaults
   :default-activation-fn :relu :default-l2 7.5e-6
   :default-weight-init :xavier :default-learning-rate 0.0015
   :default-updater :nesterovs :default-momentum 0.98

   ;; setting layer configuration
   :layers {0 {:dense-layer
               {:layer-name "example first layer"
                :n-in 784 :n-out 500}}
            1 {:dense-layer
               {:layer-name "example second layer"
                :n-in 500 :n-out 100}}
            2 {:output-layer
               {:n-in 100 :n-out 10
                ;; layer specific params
                :loss-fn :negativeloglikelihood
                :activation-fn :softmax
                :layer-name "example output layer"}}}

   ;; multi layer args
   :backprop? true
   :pretrain? false))

(def multi-layer-network (c/model-from-conf nn-conf))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; local cpu training with dl4j pre-built iterators
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; lets use the pre-built Mnist data set iterator

(def train-mnist-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? true
   :seed 123))

(def test-mnist-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? false
   :seed 123))

;; and lets set a listener so we can know how training is going

(def score-listener (listener/new-score-iteration-listener :print-every-n 5))

;; and attach it to our model

;; TODO: listeners are broken, look into log4j warnning
(def mln-with-listener (set-listeners! :model multi-layer-network
                                       :listeners [score-listener]))

(def trained-mln (mln/train-mln-with-ds-iter! :mln mln-with-listener
                                              :iter train-mnist-iter
                                              :n-epochs 15
                                              :as-code? false))

;; training happens because :as-code? = false
;; if it was true, we would still just have a data structure
;; we now have a trained model that has seen the training dataset 15 times
;; time to evaluate our model

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;;Create an evaluation object
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def eval-obj (evaluate-classification :mln trained-mln
                                       :iter test-mnist-iter))

;; always remember that these objects are stateful, dont use the same eval-obj
;; to eval two different networks
;; we trained the model on a training dataset.  We evaluate on a test set

(println (get-stats :evaler eval-obj))
;; this will print the stats to standard out for each feature/label pair

;;Examples labeled as 0 classified by model as 0: 968 times
;;Examples labeled as 0 classified by model as 1: 1 times
;;Examples labeled as 0 classified by model as 2: 1 times
;;Examples labeled as 0 classified by model as 3: 1 times
;;Examples labeled as 0 classified by model as 5: 1 times
;;Examples labeled as 0 classified by model as 6: 3 times
;;Examples labeled as 0 classified by model as 7: 1 times
;;Examples labeled as 0 classified by model as 8: 2 times
;;Examples labeled as 0 classified by model as 9: 2 times
;;Examples labeled as 1 classified by model as 1: 1126 times
;;Examples labeled as 1 classified by model as 2: 2 times
;;Examples labeled as 1 classified by model as 3: 1 times
;;Examples labeled as 1 classified by model as 5: 1 times
;;Examples labeled as 1 classified by model as 6: 2 times
;;Examples labeled as 1 classified by model as 7: 1 times
;;Examples labeled as 1 classified by model as 8: 2 times
;;Examples labeled as 2 classified by model as 0: 3 times
;;Examples labeled as 2 classified by model as 1: 2 times
;;Examples labeled as 2 classified by model as 2: 1006 times
;;Examples labeled as 2 classified by model as 3: 2 times
;;Examples labeled as 2 classified by model as 4: 3 times
;;Examples labeled as 2 classified by model as 6: 3 times
;;Examples labeled as 2 classified by model as 7: 7 times
;;Examples labeled as 2 classified by model as 8: 6 times
;;Examples labeled as 3 classified by model as 2: 4 times
;;Examples labeled as 3 classified by model as 3: 990 times
;;Examples labeled as 3 classified by model as 5: 3 times
;;Examples labeled as 3 classified by model as 7: 3 times
;;Examples labeled as 3 classified by model as 8: 3 times
;;Examples labeled as 3 classified by model as 9: 7 times
;;Examples labeled as 4 classified by model as 2: 2 times
;;Examples labeled as 4 classified by model as 3: 1 times
;;Examples labeled as 4 classified by model as 4: 967 times
;;Examples labeled as 4 classified by model as 6: 4 times
;;Examples labeled as 4 classified by model as 7: 1 times
;;Examples labeled as 4 classified by model as 9: 7 times
;;Examples labeled as 5 classified by model as 0: 2 times
;;Examples labeled as 5 classified by model as 3: 6 times
;;Examples labeled as 5 classified by model as 4: 1 times
;;Examples labeled as 5 classified by model as 5: 874 times
;;Examples labeled as 5 classified by model as 6: 3 times
;;Examples labeled as 5 classified by model as 7: 1 times
;;Examples labeled as 5 classified by model as 8: 3 times
;;Examples labeled as 5 classified by model as 9: 2 times
;;Examples labeled as 6 classified by model as 0: 4 times
;;Examples labeled as 6 classified by model as 1: 3 times
;;Examples labeled as 6 classified by model as 3: 2 times
;;Examples labeled as 6 classified by model as 4: 4 times
;;Examples labeled as 6 classified by model as 5: 4 times
;;Examples labeled as 6 classified by model as 6: 939 times
;;Examples labeled as 6 classified by model as 7: 1 times
;;Examples labeled as 6 classified by model as 8: 1 times
;;Examples labeled as 7 classified by model as 1: 7 times
;;Examples labeled as 7 classified by model as 2: 4 times
;;Examples labeled as 7 classified by model as 3: 3 times
;;Examples labeled as 7 classified by model as 7: 1005 times
;;Examples labeled as 7 classified by model as 8: 2 times
;;Examples labeled as 7 classified by model as 9: 7 times
;;Examples labeled as 8 classified by model as 0: 3 times
;;Examples labeled as 8 classified by model as 2: 3 times
;;Examples labeled as 8 classified by model as 3: 2 times
;;Examples labeled as 8 classified by model as 4: 4 times
;;Examples labeled as 8 classified by model as 5: 3 times
;;Examples labeled as 8 classified by model as 6: 2 times
;;Examples labeled as 8 classified by model as 7: 4 times
;;Examples labeled as 8 classified by model as 8: 947 times
;;Examples labeled as 8 classified by model as 9: 6 times
;;Examples labeled as 9 classified by model as 0: 2 times
;;Examples labeled as 9 classified by model as 1: 2 times
;;Examples labeled as 9 classified by model as 3: 4 times
;;Examples labeled as 9 classified by model as 4: 8 times
;;Examples labeled as 9 classified by model as 6: 1 times
;;Examples labeled as 9 classified by model as 7: 4 times
;;Examples labeled as 9 classified by model as 8: 2 times
;;Examples labeled as 9 classified by model as 9: 986 times

;;==========================Scores========================================
;; Accuracy:        0.9808
;; Precision:       0.9808
;; Recall:          0.9807
;; F1 Score:        0.9807
;;========================================================================

;; can get the stats that are printed via fns in the evaluation namespace
;; after running eval-model-whole-ds

(get-accuracy :evaler evaler-with-stats) ;; => 0.9808

Model Tuning

Early Stopping (controlling training)

it is recommened you start here when designing models

using dl4clj.core


(ns my.ns
  (:require [dl4clj.earlystopping.termination-conditions :refer :all]
            [dl4clj.earlystopping.model-saver :refer [new-in-memory-saver]]
            [dl4clj.nn.api.multi-layer-network :refer [evaluate-classification]]
            [dl4clj.eval.api.eval :refer [get-stats]]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.datasets.iterators :as iter]
            [dl4clj.core :as c]))

(def nn-conf
  (nn/builder
   ;; network args
   :optimization-algo :stochastic-gradient-descent
   :seed 123
   :iterations 1
   :regularization? true

   ;; setting layer defaults
   :default-activation-fn :relu
   :default-l2 7.5e-6
   :default-weight-init :xavier
   :default-learning-rate 0.0015
   :default-updater :nesterovs
   :default-momentum 0.98

   ;; setting layer configuration
   :layers {0 {:dense-layer
               {:layer-name "example first layer"
                :n-in 784 :n-out 500}}
            1 {:dense-layer
               {:layer-name "example second layer"
                :n-in 500 :n-out 100}}
            2 {:output-layer
               {:n-in 100 :n-out 10
                ;; layer specific params
                :loss-fn :negativeloglikelihood
                :activation-fn :softmax
                :layer-name "example output layer"}}}

   ;; multi layer args
   :backprop? true
   :pretrain? false))

(def train-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? true
   :seed 123))

(def test-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? false
   :seed 123))

(def invalid-score-condition (new-invalid-score-iteration-termination-condition))

(def max-score-condition (new-max-score-iteration-termination-condition
                          :max-score 20.0))

(def max-time-condition (new-max-time-iteration-termination-condition
                         :max-time-val 10
                         :max-time-unit :minutes))

(def score-doesnt-improve-condition (new-score-improvement-epoch-termination-condition
                                     :max-n-epoch-no-improve 5))

(def target-score-condition (new-best-score-epoch-termination-condition
                             :best-expected-score 0.009))

(def max-number-epochs-condition (new-max-epochs-termination-condition :max-n 20))

(def in-mem-saver (new-in-memory-saver))

(def trained-mln
;; defaults to returning the model
  (c/train-with-early-stopping
   :nn-conf nn-conf
   :training-iter train-mnist-iter
   :testing-iter test-mnist-iter
   :eval-every-n-epochs 1
   :iteration-termination-conditions [invalid-score-condition
                                      max-score-condition
                                      max-time-condition]
   :epoch-termination-conditions [score-doesnt-improve-condition
                                  target-score-condition
                                  max-number-epochs-condition]
   :save-last-model? true
   :model-saver in-mem-saver
   :as-code? false))

(def model-evaler
  (evaluate-classification :mln trained-mln :iter test-mnist-iter))

(println (get-stats :evaler model-evaler))
  • explicit, step by step way of doing this
(ns my.ns
  (:require [dl4clj.earlystopping.early-stopping-config :refer [new-early-stopping-config]]
            [dl4clj.earlystopping.termination-conditions :refer :all]
            [dl4clj.earlystopping.model-saver :refer [new-in-memory-saver new-local-file-model-saver]]
            [dl4clj.earlystopping.score-calc :refer [new-ds-loss-calculator]]
            [dl4clj.earlystopping.early-stopping-trainer :refer [new-early-stopping-trainer]]
            [dl4clj.earlystopping.api.early-stopping-trainer :refer [fit-trainer!]]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.nn.multilayer.multi-layer-network :as mln]
            [dl4clj.utils :refer [load-model!]]
            [dl4clj.datasets.iterators :as iter]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; start with our network config
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def nn-conf
  (nn/builder
   ;; network args
   :optimization-algo :stochastic-gradient-descent
   :seed 123 :iterations 1 :regularization? true
   ;; setting layer defaults
   :default-activation-fn :relu :default-l2 7.5e-6
   :default-weight-init :xavier :default-learning-rate 0.0015
   :default-updater :nesterovs :default-momentum 0.98
   ;; setting layer configuration
   :layers {0 {:dense-layer
               {:layer-name "example first layer"
                :n-in 784 :n-out 500}}
            1 {:dense-layer
               {:layer-name "example second layer"
                :n-in 500 :n-out 100}}
            2 {:output-layer
               {:n-in 100 :n-out 10
                ;; layer specific params
                :loss-fn :negativeloglikelihood
                :activation-fn :softmax
                :layer-name "example output layer"}}}
   ;; multi layer args
   :backprop? true
   :pretrain? false))

(def mln (c/model-from-conf nn-conf))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; the training/testing data
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def train-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? true
   :seed 123))

(def test-iter
  (iter/new-mnist-data-set-iterator
   :batch-size 64
   :train? false
   :seed 123))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; we are going to need termination conditions
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; these allow us to control when we exit training

;; this can be based off of iterations or epochs

;; iteration termination conditions

(def invalid-score-condition (new-invalid-score-iteration-termination-condition))

(def max-score-condition (new-max-score-iteration-termination-condition
                          :max-score 20.0))

(def max-time-condition (new-max-time-iteration-termination-condition
                         :max-time-val 10
                         :max-time-unit :minutes))

;; epoch termination conditions

(def score-doesnt-improve-condition (new-score-improvement-epoch-termination-condition
                                     :max-n-epoch-no-improve 5))

(def target-score-condition (new-best-score-epoch-termination-condition :best-expected-score 0.009))

(def max-number-epochs-condition (new-max-epochs-termination-condition :max-n 20))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; we also need a way to save our model
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; can be in memory or to a local directory

(def in-mem-saver (new-in-memory-saver))

(def local-file-saver (new-local-file-model-saver :directory "resources/tmp/readme/"))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; set up your score calculator
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def score-calcer (new-ds-loss-calculator :iter test-iter
                                          :average? true))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; create an early stopping configuration
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; termination conditions
;; a way to save our model
;; a way to calculate the score of our model on the dataset

(def early-stopping-conf
  (new-early-stopping-config
   :epoch-termination-conditions [score-doesnt-improve-condition
                                  target-score-condition
                                  max-number-epochs-condition]
   :iteration-termination-conditions [invalid-score-condition
                                      max-score-condition
                                      max-time-condition]
   :eval-every-n-epochs 5
   :model-saver local-file-saver
   :save-last-model? true
   :score-calculator score-calcer))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; create an early stopping trainer from our data, model and early stopping conf
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def es-trainer (new-early-stopping-trainer :early-stopping-conf early-stopping-conf
                                            :mln mln
                                            :iter train-iter))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; fit and use our early stopping trainer
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def es-trainer-fitted (fit-trainer! es-trainer :as-code? false))

;; when the trainer terminates, you will see something like this
;;[nREPL-worker-24] BaseEarlyStoppingTrainer INFO  Completed training epoch 14
;;[nREPL-worker-24] BaseEarlyStoppingTrainer INFO  New best model: score = 0.005225599372851298,
;;                                                   epoch = 14 (previous: score = 0.018243224899038346, epoch = 7)
;;[nREPL-worker-24] BaseEarlyStoppingTrainer INFO Hit epoch termination condition at epoch 14.
;;                                           Details: BestScoreEpochTerminationCondition(0.009)

;; and if we look at the es-trainer-fitted object we see

;;#object[org.deeplearning4j.earlystopping.EarlyStoppingResult 0x5ab74f27 EarlyStoppingResult
;;(terminationReason=EpochTerminationCondition,details=BestScoreEpochTerminationCondition(0.009),
;; bestModelEpoch=14,bestModelScore=0.005225599372851298,totalEpochs=15)]

;; and our model has been saved to /resources/tmp/readme/bestModel.bin
;; there we have our model config, model params and our updater state

;; we can then load this model to use it or continue refining it

(def loaded-model (load-model! :path "resources/tmp/readme/bestModel.bin"
                               :load-updater? true))

Transfer Learning (freezing layers)


;; TODO: need to write up examples

Spark Training

dl4j Spark usage

How it is done in dl4clj

  • Uses dl4clj.core
    • This example uses a fn which takes care of most steps for you
      • allows you to pass args as code bc the fn accounts for the multiple spark contexts issue encountered when everything is just a data structure

(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.datasets.iterators :refer [new-iris-data-set-iterator]]
            [dl4clj.eval.api.eval :refer [get-stats]]
            [dl4clj.spark.masters.param-avg :as master]
            [dl4clj.spark.data.java-rdd :refer [new-java-spark-context
                                                java-rdd-from-iter]]
            [dl4clj.spark.api.dl4j-multi-layer :refer [eval-classification-spark-mln
                                                       get-spark-context]]
            [dl4clj.core :as c]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 1, create your model config
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def mln-conf
  (nn/builder
   :optimization-algo :stochastic-gradient-descent
   :default-learning-rate 0.006
   :layers {0 (l/dense-layer-builder :n-in 4 :n-out 2 :activation-fn :relu)
            1 {:output-layer
               {:loss-fn :negativeloglikelihood
                :n-in 2 :n-out 3
                :activation-fn :soft-max
                :weight-init :xavier}}}
   :backprop? true
   :backprop-type :standard))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 2, training master
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def training-master
  (master/new-parameter-averaging-training-master
   :build? true
   :rdd-n-examples 10
   :n-workers 4
   :averaging-freq 10
   :batch-size-per-worker 2
   :export-dir "resources/spark/master/"
   :rdd-training-approach :direct
   :repartition-data :always
   :repartition-strategy :balanced
   :seed 1234
   :save-updater? true
   :storage-level :none))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 3, spark context
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def your-spark-context
  (new-java-spark-context :app-name "example app"))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 4, training data
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def iris-iter
  (new-iris-data-set-iterator
   :batch-size 1
   :n-examples 5))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 5, spark mln
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def fitted-spark-mln
  (c/train-with-spark :spark-context your-spark-context
                      :mln-conf mln-conf
                      :training-master training-master
                      :iter iris-iter
                      :n-epochs 1
                      :as-code? false))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 5, use spark context from spark-mln to create rdd
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; TODO: eliminate this step

(def our-rdd
  (let [sc (get-spark-context fitted-spark-mln :as-code? false)]
    (java-rdd-from-iter :spark-context sc
                        :iter iris-iter)))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 6, evaluation model and print stats (poor performance of model expected)
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def eval-obj
  (eval-classification-spark-mln
   :spark-mln fitted-spark-mln
   :rdd our-rdd))

(println (get-stats :evaler eval-obj))

  • this example demonstrates the dl4j workflow
    • NOTE: unlike the previous example, this one requires dl4j objects to be used
      • this is becaues spark only wants you to have one spark context at a time
(ns my.ns
  (:require [dl4clj.nn.conf.builders.layers :as l]
            [dl4clj.nn.conf.builders.nn :as nn]
            [dl4clj.datasets.iterators :refer [new-iris-data-set-iterator]]
            [dl4clj.eval.api.eval :refer [get-stats]]
            [dl4clj.spark.masters.param-avg :as master]
            [dl4clj.spark.data.java-rdd :refer [new-java-spark-context java-rdd-from-iter]]
            [dl4clj.spark.dl4j-multi-layer :as spark-mln]
            [dl4clj.spark.api.dl4j-multi-layer :refer [fit-spark-mln!
                                                       eval-classification-spark-mln]]))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 1, create your model
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def mln-conf
  (nn/builder
   :optimization-algo :stochastic-gradient-descent
   :default-learning-rate 0.006
   :layers {0 (l/dense-layer-builder :n-in 4 :n-out 2 :activation-fn :relu)
            1 {:output-layer
               {:loss-fn :negativeloglikelihood
                :n-in 2 :n-out 3
                :activation-fn :soft-max
                :weight-init :xavier}}}
   :backprop? true
   :as-code? false
   :backprop-type :standard))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 2, create a training master
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; not all options specified, but most are

(def training-master
  (master/new-parameter-averaging-training-master
   :build? true
   :rdd-n-examples 10
   :n-workers 4
   :averaging-freq 10
   :batch-size-per-worker 2
   :export-dir "resources/spark/master/"
   :rdd-training-approach :direct
   :repartition-data :always
   :repartition-strategy :balanced
   :seed 1234
   :as-code? false
   :save-updater? true
   :storage-level :none))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 3, create a Spark Multi Layer Network
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def your-spark-context
  (new-java-spark-context :app-name "example app" :as-code? false))

;; new-java-spark-context will turn an existing spark-configuration into a java spark context
;; or create a new java spark context with master set to "local[*]" and the app name
;; set to :app-name


(def spark-mln
  (spark-mln/new-spark-multi-layer-network
   :spark-context your-spark-context
   :mln mln-conf
   :training-master training-master
   :as-code? false))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 4, load your data
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

;; one way is via a dataset-iterator
;; can make one directly from a dataset (iterator data-set)
;; see: nd4clj.linalg.dataset.api.data-set and nd4clj.linalg.dataset.data-set
;; we are going to use a pre-built one

(def iris-iter
  (new-iris-data-set-iterator
   :batch-size 1
   :n-examples 5
   :as-code? false))

;; now lets convert the data into a javaRDD

(def our-rdd
  (java-rdd-from-iter :spark-context your-spark-context
                      :iter iris-iter))

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
;; Step 5, fit and evaluate the model
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

(def fitted-spark-mln
  (fit-spark-mln!
   :spark-mln spark-mln
   :rdd our-rdd
   :n-epochs 1))
;; this fn also has the option to supply :path-to-data instead of :rdd
;; that path should point to a directory containing a number of dataset objects

(def eval-obj
  (eval-classification-spark-mln
   :spark-mln fitted-spark-mln
   :rdd our-rdd))
;; we would want to have different testing and training rdd's but here we are using
;; the data we trained on

;; lets get the stats for how our model performed

(println (get-stats :evaler eval-obj))

Terminology

Coming soon

Packages to come back to:

Implement ComputationGraphs and the classes which use them

NLP

Parallelism

TSNE

UI


Author: yetanalytics
Source Code: https://github.com/yetanalytics/dl4clj
License: BSD-2-Clause License

#machine-learning #deep-learning 

Arvel  Parker

Arvel Parker

1591611780

How to Find Ulimit For user on Linux

How can I find the correct ulimit values for a user account or process on Linux systems?

For proper operation, we must ensure that the correct ulimit values set after installing various software. The Linux system provides means of restricting the number of resources that can be used. Limits set for each Linux user account. However, system limits are applied separately to each process that is running for that user too. For example, if certain thresholds are too low, the system might not be able to server web pages using Nginx/Apache or PHP/Python app. System resource limits viewed or set with the NA command. Let us see how to use the ulimit that provides control over the resources available to the shell and processes.

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]

MEAN Stack Tutorial MongoDB ExpressJS AngularJS NodeJS

We are going to build a full stack Todo App using the MEAN (MongoDB, ExpressJS, AngularJS and NodeJS). This is the last part of three-post series tutorial.

MEAN Stack tutorial series:

AngularJS tutorial for beginners (Part I)
Creating RESTful APIs with NodeJS and MongoDB Tutorial (Part II)
MEAN Stack Tutorial: MongoDB, ExpressJS, AngularJS and NodeJS (Part III) 👈 you are here
Before completing the app, let’s cover some background about the this stack. If you rather jump to the hands-on part click here to get started.

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]

CentOS Linux 8.2 Released and Here is How to Upgrade it

CentOS Linux 8.2 (2004) released. It is a Linux distribution derived from RHEL (Red Hat Enterprise Linux) 8.2 source code. CentOS was created when Red Hat stopped providing RHEL free. CentOS 8.2 gives complete control of its open-source software packages and is fully customized for research needs or for running a high-performance website without the need for license fees. Let us see what’s new in CentOS 8.2 (2004) and how to upgrade existing CentOS 8.1.1199 server to 8.2.2004 using the command line.

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]

Brain  Crist

Brain Crist

1595434320

Docker Applikationen mit Visual Studio Code debuggen

Mit dem integrierten Debugger von Visual Studio Code lassen sich ASP.NET Core bzw. .NET Core Applikationen einfach und problemlos debuggen. Der Debugger unterstützt auch Remote Debugging, somit lassen sich zum Beispiel .NET Core Programme, die in einem Docker-Container laufen, debuggen.

Als Beispiel Applikation reicht das Default-Template für MVC Applikationen dotnet new mvc

$ md docker-core-debugger
$ cd docker-core-debugger
$ dotnet new mvc

Mit dotnet run prüfen wir kurz, ob die Applikation läuft und unter der Adresse http://localhost:5000 erreichbar ist.

$ dotnet run
$ Hosting environment: Production
$ Content root path: D:\Temp\docker-aspnetcore
$ Now listening on: http://localhost:5000

Die .NET Core Applikation builden wir mit dotnet build und publishen alles mit Hilfe von dotnet publish

$ dotnet build
$ dotnet publish -c Debug -o out --runtime linux-x64

Dabei gilt es zu beachten, dass die Build Configuration mit -c Debug gesetzt ist und das Output Directory auf -o out. Sonst findet Docker die nötigen Binaries nicht. Für den Docker Container brauchen wir nun ein Dockerfile, dass beim Start vorgängig den .NET Core command line debugger (VSDBG) installiert. Das Installations-Script für VSDBG ist unter https://aka.ms/getvsdbgsh abfrufbar.

FROM microsoft/aspnetcore:latest
WORKDIR /app

RUN apt-get update \
    && apt-get install -y --no-install-recommends \
       unzip procps \
    && rm -rf /var/lib/apt/lists/* \
    && curl -sSL https://aka.ms/getvsdbgsh | bash /dev/stdin -v latest -l /vsdbg

COPY ./out .
ENTRYPOINT ["dotnet", "docker-core-debugger.dll"]

Den Docker Container erstellen wir mit dem docker build Kommando

$ docker build -t coreapp .

und starten die Applikation mit docker run.

$ docker run -d -p 8080:80 --name coreapp coreapp

Jetzt muss Visual Studio Code nur noch wissen, wo unsere Applikation läuft. Dazu definieren wir eine launch.json vom Typ attach und konfigurieren die nötigen Parameter für den Debugger.

{
    "version": "0.2.0",
    "configurations": [
         {
            "name": ".NET Core Remote Attach",
            "type": "coreclr",
            "request": "attach",
            "processId": "${command:pickRemoteProcess}",
            "pipeTransport": {
                "pipeProgram": "docker",
                "pipeArgs": ["exec", "-i coreapp ${debuggerCommand}"],
                "quoteArgs": false,
                "debuggerPath": "/vsdbg/vsdbg",
                "pipeCwd": "${workspaceRoot}"
            },

            "logging": {
                "engineLogging": true,
                "exceptions": true,
                "moduleLoad": true,
                "programOutput": true
            },
        }
    ]
}

Mit F5 starten wir den Debugger. Wenn alles klappt, sollte eine Auswahl der Prozesse des Docker-Containers sichtbar sein.

vscode

Nun muss der dotnet Prozess ausgewählt werden. Der Visual Studio Code Debugger verbindet sich darauf mit VSDBG und wir können wie gewohnt unseren Code debuggen. Dazu setzen wir einen Breakpoint in der Index-Action des HomeControllers und rufen mit dem Browser die URL http://localhost:8080/ auf.

vscode

#[object object] #[object object] #[object object] #[object object] #[object object] #[object object] #[object object]